
Fix the ChatGPT em dash problem. A quick guide to update your custom instructions and control AI style.

You can stop ChatGPT from using em dashes by updating your personalization settings. Follow these three steps:
The ChatGPT em dash problem was the AI's constant and excessive use of the em dash (—) punctuation mark. This occurred even when users explicitly asked it not to.
This writing habit became a frustrating and well-known trait of ChatGPT's output. Users on platforms like Reddit, Twitter, and OpenAI's own forums shared widespread complaints about the issue.
The em dash became so common that many people started using it to spot AI-generated content. Its frequent use in school papers, work emails, and ad copy made it an unofficial signature of ChatGPT.
Some users even reported being accused of using AI simply for using em dashes in their own writing. This created a problem for writers who needed to produce content that felt human and authentic.
The model's training data was the source of the problem. ChatGPT was trained on vast amounts of text from academic papers, journalism, and literature, where em dashes are common.
This deep-seated pattern in the training data caused the AI to favor the punctuation, often overriding direct user commands to avoid it. Users who were paying for the service found this inability to control the output particularly frustrating.
OpenAI acknowledged the user feedback and released an update to address the problem. The fix gives you more reliable control over the AI's output style. As the original article from TechCrunch AI explains, the model now properly respects your preferences.
This change does not remove the em dash by default. Instead, it empowers you to stop its use by providing a clear instruction. OpenAI CEO Sam Altman confirmed the update, calling it a "small-but-happy win."
Updating your settings is simple and takes less than a minute. Here is the exact process to follow:
The fix for the ChatGPT em dash is part of a larger update that gives you more granular control over the AI's personality and tone. Announced on November 12, 2025, with the GPT-5.1 model, these features help you align the AI's output with your specific needs.
You can now adjust response traits like warmth, conciseness, scannability, and even emoji use. This helps ensure every piece of content the AI generates matches your brand voice from the start, saving significant editing time.
OpenAI introduced eight preset chat styles. You can select one in the personalization tab to instantly change the tone of all responses. The available styles are:
These new controls are essential for businesses. When you're creating articles, blog posts, or website copy, brand consistency is key. A uniform tone and style build trust and make your content recognizable.
By defining your writing style directly in ChatGPT, you create a powerful assistant for generating high-quality drafts. This helps your team produce better SEO content faster, all while staying perfectly on-brand.
This update is a practical step toward making AI a more reliable tool for professional creators. You can spend less time editing out awkward punctuation and more time focusing on strategy and ideas.
The ability to set a consistent, brand-aligned tone is crucial for managing your digital presence. It ensures that every blog article, email, and social media post sounds like it came from you. This level of control helps streamline workflows for agencies and marketing teams managing multiple client voices.
For example, you can now generate a month's worth of on-brand posts without manually correcting the style of each one. This makes effective social media management more efficient and scalable.



No guesswork, no back-and-forth. Just one team managing your website, content, and social. Built to bring in traffic and results.